Pattern classification of endocervical adenocarcinoma: reproducibility and review of criteria.

Previously, our international team proposed a three-tiered pattern classification (Pattern Classification) system for endocervical adenocarcinoma of the usual type that correlates with nodal disease and recurrence. Pattern Classification-A tumors have well-demarcated glands lacking destructive stromal invasion or lymphovascular invasion, Pattern Classification-B tumors show localized, limited destructive invasion arising from A-type glands, and Pattern Classification-C tumors have diffuse destructive stromal invasion, significant (filling a 4 × field) confluence, or solid architecture. Twenty-four cases of Pattern Classification-A, 22 Pattern Classification-B, and 38 Pattern Classification-C from the tumor set used in the original description were chosen using the reference diagnosis originally established. One H&E slide per case was reviewed by seven gynecologic pathologists, four from the original study. Kappa statistics were prepared, and cases with discrepancies reviewed. We found a majority agreement with reference diagnosis in 81% of cases, with complete or near-complete (six of seven) agreement in 50%. Overall concordance was 74%. Overall kappa (agreement among pathologists) was 0.488 (moderate agreement). Pattern Classification-B has lowest kappa, and agreement was not improved by combining B+C. Six of seven reviewers had substantial agreement by weighted kappas (>0.6), with one reviewer accounting for the majority of cases under or overcalled by two tiers. Confluence filling a 4 × field, labyrinthine glands, or solid architecture accounted for undercalling other reference diagnosis-C cases. Missing a few individually infiltrative cells was the most common cause of undercalling reference diagnosis-B. Small foci of inflamed, loose or desmoplastic stroma lacking infiltrative tumor cells in reference diagnosis-A appeared to account for those cases up-graded to Pattern Classification-B. In summary, an overall concordance of 74% indicates that the criteria can be reproducibly applied by gynecologic pathologists. Further refinement of criteria should allow use of this powerful classification system to delineate which cervical adenocarcinomas can be safely treated conservatively.